Oral Oncology 47 (2011) 487–496

Contents lists available at ScienceDirect

Oral Oncology

journal homepage: www.elsevier .com/locate /ora loncology

Clinical significance of genetic alterations and expression of epidermal growthfactor receptor (EGFR) in head and neck squamous cell carcinomas

Balázs Szabó a, Györgyi A. Nelh}ubel a,b, Adél Kárpáti a,b, István Kenessey c, Balázs Jóri d, Csilla Székely d,István Peták d, Gábor Lotz c, Zita Heged}us c, Balázs Heged}us c,e, Tibor Füle f, Balázs Döme e,g, József Tímár c,József Tóvári b,g,⇑a Department of Otolaryngology and Head and Neck Surgery, Semmelweis University, Budapest, Hungaryb Department of Experimental Pharmacology, National Institute of Oncology, Budapest, Hungaryc 2nd Department of Pathology, Semmelweis University, Budapest, Hungaryd KPS Medical Biotechnology and Healthcare Services Ltd., Budapest, Hungarye Department of Thoracic Surgery, Medical University of Vienna, Vienna, Austriaf 1st Institute of Pathology and Experimental Cancer Research, Semmelweis University, Budapest, Hungaryg Department of Tumor Biology, National Korányi Institute of Pulmonology, Budapest, Hungary

a r t i c l e i n f o s u m m a r y

Article history:Received 30 November 2010Received in revised form 8 March 2011Accepted 14 March 2011Available online 16 April 2011

Keywords:EGFR gene alterationsEGFR protein expressionHPVHead and neck cancer

1368-8375/$ - see front matter � 2011 Elsevier Ltd. Adoi:10.1016/j.oraloncology.2011.03.020

⇑ Corresponding author at: Department of ExperimInstitute of Oncology, Budapest, Hungary. Tel.: +36 1 2

E-mail addresses: jtovari@yahoo.com, tozsi@oncol

The significance of epidermal growth factor receptor (EGFR) signaling is well studied in a number of dif-ferent tumors, but limited data is available with regard to head and neck squamous cell carcinoma(HNSCC). Since anti-EGFR therapies are currently under investigation in these malignancies as well, com-prehensive information about the alteration of EGFR in HNSCC is necessary to design the most appropri-ate therapeutic protocols.

We examined retrospectively the gene copy number of EGFR by FISH and the protein expression byimmunohistochemistry using different epitope-specific antibodies in paraffin-embedded primary tumorsof five different regions, from 71 HNSCC patients who had not been treated with anti-EGFR therapy. Inseven cases corresponding lymph node metastases were also available for comparative analyses. We alsodetermined the mutational status of tyrosine kinase (TK) domain (exon 19 and 21) and the extracellulardeletion mutation (vIII) of EGFR, the KRAS mutation at codon 12 and the presence of HPV infection.

Eight of the 71 cases (11.3%) showed EGFR gene amplification (most of them localized into the hypo-pharyngeal region) and the increased gene copy number (amplification + polysomy) was 43.7%. Despitepronounced intratumoral heterogeneity of EGFR protein expression being found, the high EGFR expres-sion correlated with poor prognosis. On the other hand, the phosphorylation of EGFR was associated withprolonged survival. No mutations in the TK domain of EGFR were found in any of the HNSCC patients andonly two cases were KRAS mutant at codon 12. We detected vIII deletion mutation of EGFR in 21% of thesamples, but there was no statistically significant correlation between the presence of vIII mutant formand patient survival. EGFR vIII mutation was, however, associated with increased gene copy number.Fourteen of 71 cases (19.7%) were HPV-positive and the incidence of infection showed a decreasing ten-dency from the oral cavity towards the larynx. Interestingly, in contrast to previous findings, we could notobserve improved survival in HPV-positive patients compared to non-infected patients, most probablydue to the fact that the majority of these HNSCC patients were smokers and alcohol consumers.

In conclusion, we found that increased EGFR protein levels and gene copy numbers (not gene amplifi-cation alone) have prognostic significance in the investigated HNSCC patient population. However, therelatively high incidence of the EGFR-vIII mutant form warrants careful therapeutic decision-makingwhen choosing between different anti-EGFR treatment options.

� 2011 Elsevier Ltd. All rights reserved.

ll rights reserved.

ental Pharmacology, National24 8778; fax: +36 1 224 8724..hu (J. Tóvári).

Introduction

Head and neck squamous cell carcinoma (HNSCC) is the sixthmost frequent tumor type with approximately 600,000 new caseseach year.1 In Hungary HNSCC is the third most frequent type ofmalignant tumor among males, and Hungary is ranked first in

Table 1Pathological data of patients.

Variables n

GenderFemale 13Male 58Age (at the time of the diagnosis)<54 34P54 37Tumor siteTonsil 10Base of tongue 6Glottic 20Supraglottic 16Hypopharynx 19Metastasis 7Alcohol consumptionStrong 27Moderate 13Occasional/never 29Past consumer 2SmokingYes/strong 67No/moderate 2Already quit 2StageI 8II 32III 19IV 10Unknown 2Tumor gradeI 19II 42III 10N stageN0 30N+ 29Nx 12Surgical marginTumor free/negative 43Positive 28Vessel invasionNegative 32Positive 39Inflammational infiltrationNo/mild 9Medium/strong 62

488 B. Szabó et al. / Oral Oncology 47 (2011) 487–496

the world with respect to its incidence and related death occur-rences.2 Early-stage (I–II) tumors are treated with surgery or radi-ation therapy as monotherapy, while advanced tumors needcombined treatment, surgical intervention and chemo-radiother-apy.3 Currently the 5-year survival rate for stage III–IV tumors var-ies between 30% and 40% in most clinical trials.4 In order toincrease the survival rate, new treatment modalities should beintroduced based on the molecular characteristics of HNSCCs.

EGFR signaling is a major regulator of proliferation and differen-tiation in a number of organ systems during development and inthe maintenance of tissue homeostasis. In tumor cells, however,it contributes to proliferation and apoptosis inhibition as well asinvasion, adhesion and metastasis formation.3,5 The EGF receptoris often overexpressed in the majority of tumors of epithelial ori-gin: in non-small cell lung cancer, colorectal carcinomas, in ovar-ian, breast, prostate and renal cancer, as well as in HNSCC.3,6

Currently EGFR signaling is targeted pharmacologically in twoways: the ligand-binding region of the receptor can be inhibitedby monoclonal antibodies or, alternatively, the receptor autophos-phorylation can be hampered by low-molecular-weight tyrosinekinase inhibitors (TKIs), thus inhibiting the activation of signaltransduction pathways. Both of these strategies are used as partof combination therapies. However, alterations in EGFR expressionand its mutations significantly influence these targeted therapeuticapproaches.

EGFR protein overexpression has been reported in 70–90% ofhead and neck tumors in various studies.4,7,8 Furthermore theincidence of gene amplification has been demonstrated in about17–31% of head and neck tumors.9,10 Importantly, EGFR overex-pression and amplification has been associated with a worseprognosis.7,10

A major oncogenic mutation of the extracellular domain resultsin the vIII variant that has been identified in several types of malig-nancies. The presence of this mutation in HNSCC is described inonly one study, where 42% of the cases were found to be expressingthis mutated protein.11

The mutations in exons 18, 19 and 21 of the TK domain of EGFRhave a role in predicting the therapeutic success of TKI treatmentin lung cancer. Accordingly, several studies tried to detect thesemutations in head and neck cancer as well.12,13 Based on previousstudies, the frequency of TK domain mutation in HNSCC is verylow. It has been demonstrated to be present in a small percentageof cases in Asian populations, while in European studies it has notbeen verified in any cases.14–16

RAS proteins are critical regulators of EGFR signaling. Oncogenicmutation in HRAS is a rather frequent genetic alteration in certainAsian populations, but its incidence in Europe is very low.8,17 Not-withstanding, KRAS mutations are predictive factors for anti-EGFRtherapies in certain malignancies, thus we sought to determine theincidence of KRAS mutations in our patients.

Besides the known factors in the etiology of head and neck can-cer, namely smoking and alcohol consumption, there are studiessuggesting a potential role of HPV infection. Recently publishedstudies showed the presence of the viral genes in 19–35% of HNSCCcases.18,19 According to a meta-analysis published in 2005, theprevalence of HPV infection is 25.9% worldwide.20 The most fre-quent oncogenic HPV strains are 16 and 18, just like in the caseof cervical cancers.19,21 Since previous studies have suggested thatHPV related HNSSC cases follow a different course of disease wealso determined the presence of HPV infection.18

Accordingly, in the present study we investigated the expres-sion level of the EGFR protein in 71 HNSCC patients, using four dif-ferent epitope-specific anti-EGFR antibodies, and compared it tothat of normal peritumoral epithelium. In addition, we studiedknown EGFR gene alterations (amplification, vIII mutation, TK19–21 mutations), mutation of KRAS at codon 12, and incidence

of HPV (E6 ORF sequence) in our samples and analyzed their prog-nostic significance. Furthermore, we compared the molecular char-acteristics in seven pairs of primary tumors and correspondinglymph node metastases.

Materials and methods

Patients

Surgical specimens of 71 HNSCC patients who had been treatedat Semmelweis University, Department of Otorhinolaryngologyand Head and Neck Surgery, in the period of 2001–2006 were stud-ied. Primary tumor samples were collected from the followingregions: 19 from the hypopharynx, six from the base of the tongue,10 from the tonsils, 20 from the glottis and 16 from the supraglot-tic region. In seven cases we also studied primary tumor-associated locoregional lymphatic metastases. Pathological dataof the patients are summarized in Table 1.

Tissue microarray

For the tissue microarray (TMA), hematoxylin and eosin-stainedsections were used to define the tumor areas and two representa-tive 0.6 mm cores were obtained from each case and inserted in a

B. Szabó et al. / Oral Oncology 47 (2011) 487–496 489

grid pattern into a recipient paraffin block using a tissue arrayer(3D Histech, Budapest, Hungary). Sections (2 lm) were then cutfrom each TMA block and stained with antibodies or used in FISHmethod.

Immunohistochemistry

Sections were deparaffinized in xylene for 2 � 20 min, then inethanol for 2 � 15 min. Endogenous peroxidase activity wasblocked with methanol and hydrogen peroxide for 20 min, andcontinued with 3 � 5 min washing with distilled water. Microwaveantigen retrieval (MFX-800-3 automatic microwave device, 750 W,Meditest, Budapest, Hungary) was performed at 95 �C for10 + 5 min in citrate buffer (0.05 mM, pH 6, DakoCytomation,Glostrup, Denmark). Sections were blocked for 20 min in 3% bovineserum albumin (Sigma–Aldrich, St. Louis, MO) at room tempera-ture, and then washed for 5 min in TRIS buffer. The epitope-specificantibodies were as follows: monoclonal mouse antibody recogniz-ing the outer extracellular region close to ligand-binding domain(IgG1, clone E30, DakoCytomation, dilution 1:40); monoclonalmouse antibody against the membrane-proximal extracellular re-gion (IgG2, NCL-EGFR, Novocastra Laboratories, Newcastle uponTyne, UK, dilution 1:40); polyclonal rabbit antibody against theintracellular domain of EGFR (PU335-UP, BioGenex, San Remon,CA, dilution 1:10); phospho-specific polyclonal rabbit antibodybinding to the phospho-tyrosine autophosphorylation site 1086on the tyrosine kinase domain (44–790, Biosource InternationalInc, Camarillo, CA, dilution 1:100). EGFRvIII mutation was detectedwith mouse monoclonal antibody L8A4 (a kind gift of Darell D. Big-ner, Department of Neurology, Duke University, Durham, NC). Thespecificity of the antibody for the junction of the fusion of exonsone to eight found in EGFRvIII was described previously.11,22

Negative controls were prepared with the omission of primaryantibody. Primary antibodies had been applied overnight in awet-chamber at 5 �C. As secondary antibody we used biotin-conju-gated anti-mouse/anti-rabbit IgG antibody (LSAB2 System kit, Dak-oCytomation) for 10 min followed by washing for 3 � 5 min, andincubation with streptavidin-horseradish peroxidase for 10 min.Reaction was developed with diaminobenzidine after washing for3 � 5 min. Nuclear staining was performed with hematoxilin(DakoCytomation) for 20 s, then sections were covered using glyc-erol-gelatine and examined under a light microscope.

Immunohistochemical assessment

Sections were examined at 400� magnification and stainingintensity was evaluated using a 4� scale. Staining of the normal epi-thelium served as a reference value for 2+. Tumor areas thatstained less intensely than the epithelial basal layer were consid-ered as 1+. Areas with more intense staining were considered as3+, while no staining was graded zero. We applied this scaling in-stead of the EGFR PharmDx™ classification because the stainingwas highly variable within the same tumor area. Using a 20�objective lens, the percentages of cells with different stainingintensities were determined in six fields of view.

In the case of the phospho-EGFR-specific antibody, we set thelow and high category threshold at 25% of the tumor cells positiv-ity, while in the case of the EGFRvIII-specific antibody the samplewas considered positive if the cell membrane of the tumor cellswas stained. All assessments were independently performed bytwo investigators.

Detection of EGFR gene amplification by FISH

Five micrometer thick sections were cut from paraffin-embed-ded tissue microarray blocks followed by overnight incubation at

56 �C. Slides were deparaffinized in xylene for 2 � 10 min, andrehydrated in a series of decreasing concentrations of ethanol. Heatpretreatment was carried out by microwaving the slides in VectorAntigen Unmasking Solution (H-3300; Vector Laboratories, Burlin-game, CA) at 800 W for 3 min and at 160 W for a further 20 min.Tissue sections were then digested with 0.025% pepsin (P7012; Sig-ma–Aldrich) in 0.2 M HCl at 37 �C for 15 min. Pepsin digestion wasterminated by washing the slides in distilled water for 1 min and in2� SSC for 5 min followed by dehydration in a series of increasingconcentrations of ethanol. Ten microliters of ON EGFR, Her-1(7p11)/SE 7, dual-color FISH probe mix (KBI-10702, Kreatech Bio-technology B.V., Amsterdam, The Netherlands) was applied ontothe selected area on each slide and covered with a glass coverslipwhich was sealed with rubber cement. The slides were incubatedat 80 �C for 5 min for codenaturation of sample and probe DNA,and hybridization was carried out by overnight incubation at37 �C. Posthybridization washes were performed in 0.4� SSC/0.3%Igepal for 2 min at room temperature and in 0.4� SSC/0.3% Igepalfor 2 min at 70 �C. After dehydration in a series of ethanol, Vecta-shield mounting medium with 40,6-diamidino-2-phenylindole(1.5 lg/ml, Vector Laboratories) was applied for fluorescent nucle-ar counterstaining. Leica DM RXA motorized epifluorescencemicroscope (Leica Microsystems, Wetzlar, Germany) equippedwith Pieper FK-7512-IQ high sensitivity B&W video camera andLeica CW4000 FISH visualizing software was used for the analysisand documentation of the FISH results. EGFR gene copy and thechromosome 7 numbers were counted in 40 cells from eachsample.

We considered samples as displaying polysomy or trisomy if thechromosome number was increased in more than 25% of the tumorcells.

Detection of EGFR TK domain mutations

We examined the EGFR mutational status of the samples usinga method set with the following human cell lines: EGFR wild-typeH358 bronchoalveolar carcinoma; 19del mutant HCC827 epithelialadenocarcinoma; H1975 human pulmonary adenocarcinoma witha point mutation in exon 21. The control samples contained wild-type DNA only, or wild-type and mutant DNA at a ratio of 5%, 17.5%and 50%. After normalization we were able to differentiate samplescontaining wild-type from low-concentration mutant alleles. Dur-ing the experiment we used the Roche LightCycler� 480 Real TimePCR System, adding 10 ll High Resolution Melting (HRM) Master(Roche), 0.5–0.5 ll, 10 lM forward- and reverse primers, 2.5 ll,25 mM MgCl2 and 25 ng DNA template to the 20 ll total volumereaction mixture. We used the following primers to investigatethe mutational status of exon 19 and 21 of EGFR gene, respectively:19HR01F CTGGATCCCAGAAGGTGAGA; 19HR01R GATTTCCTTGTT-GGCTTTCG; 21HR01F AGCCAGGAACGTACTGGTGA; 21HR01R TGC-CTCCTTCTGCATGGTAT. PCR conditions were as follows: 95 �C for5 min, 50 cycles of 95 �C for 15 s, 61 �C > 53 �C (decreasing by0.6 �C with each cycle) for 20 s, 72 �C for 15 s; for HRM analysis:95 �C for 1 min, 45 �C for 1 min and 61 �C > 95 �C for 25 s; finally,30 �C for 30 s.

Detection of HPV infection by PCR

For the isolation of DNA, 4–5 pieces of 10 lm sections from eachsample were placed in sterile 2 ml Eppendorf tubes. Paraffin wasremoved by two washes in xylene followed by two washes in96% ethanol. Air-dried samples were digested overnight with1 mg/ml proteinase K in Tris–EDTA buffer (10 mM Tris, 1 mM EDTApH 8.0). After high-speed centrifugation, the aqueous phase wasseparated and used as template solution. Each sample was assayedat least twice. The integrity of the DNA extracted from the archived

490 B. Szabó et al. / Oral Oncology 47 (2011) 487–496

tissues and the absence of inhibitors were controlled by subjectingthe samples to PCR amplification with GH20 and PCO4 primers tar-geted to the human beta-globin gene.

Type-specific determination of HPVs was carried out by nestedPCR. In the first round of nested PCR, a pair of outer primers de-signed for a consensus sequence shared by the three most commonhigh-risk HPV types (16, 18 and 33) was used to amplify the E6 ORFsequence. One-microliter aliquots of the amplified products fromthe first round were used as templates for the next polymerasechain reaction. In this second round, primers specific for the HPVtypes 16, 18 and 33 were applied. Amplifications were performedwith REDTaq ReadyMix (Sigma) in a PCR Express thermal cycler(Hybaid Ltd.-Thermo Electron Co.). A PCR reaction volume of25 ll was chosen, containing 12.5 ll of 2� ReadyMix, 20 pmol ofeach primer and 1 ll of template solution. Amplicons wereanalyzed on 2% DNA-quality agarose gels (Pharmacia, Uppsala,Sweden) containing 0.5 lg/ml ethidium bromide in Tris–acetate-EDTA buffer (TAE, pH 7.5), running at 5 V/cm (90 mA). PCRproducts were visualized and photographed using Kodak ImageStation 4000MM gel documentation system (Carestream HealthInc., Rochester, NY).

In cases when none of the three HPV types could be found bythe above method, an additional nested PCR assay using GP5+/GP6+ and MY9/MY11 primer pairs, targeted against L1 open read-ing frame, was attempted to detect the presence of HPV genome.These universal primer pairs recognize about 30 additional HPVtypes besides 16, 18 and 33. Detailed information about the sizeand localization of amplicons and primer sequences have been de-scribed previously.23

Detection of KRAS mutations

We used restriction fragment capillary electrophoresis to detectKRAS codon 12 mutations. First, genomic DNA was isolated fromsections of paraffin-embedded tumors using QIAamp DNA FFPETissue Kit (Qiagen, Hilden, Germany). PCR was performed withAmpliTaqGold PCR Master Mix (Applied Biosystems, Foster City,CA) using primers designed to introduce a BstnI restriction site inthe products generated from the wild-type allele. The primersequences were as follows: forward 50-GAATATAAACTTGTGG-TAGTTGGACCT-30 and reverse 50-GGTCCTGCACCAGTAATATG-30.PCR conditions were 95 �C for 10 min, 38 cycles of 95 �C for1 min, 55 �C for 1 min, 72 �C for 2 min, and finally 4 min at 72 �C.The products obtained were then digested with BstnI (New Eng-land Biolabs, Beverly, MA) at 60 �C for 4 h. This restriction enzymedigests the wild-type sequence but not the mutant one, thus iden-tifying mutations in codon 12. All products were visualized on 3%agarose gels stained with ethidium bromide. We quantified theratio of the mutant and wild-type products using Experion Auto-mated Electrophoresis System (Bio-Rad Laboratories, Hercules,CA). The transitions in the mutant alleles were determined by di-rect sequence analysis using an Applied Biosystem 3130 Geneticanalyzer (Applied Biosystems).

Statistical analysis

Categorical data were compared using Fishers’ exact probabilityand chi-square tests. Overall survival analyses were done using theKaplan–Meier method. Overall survival intervals were determinedas the time period from initial diagnosis to the time of death. Thecomparison between survival functions for different strata was as-sessed with the log-rank statistics. Multivariate analysis of prog-nostic factors was done using Cox’s regression model. Statisticalsignificance was confirmed when p values were <0.05. Statisticalanalysis was performed using Statistica 8.0 software (StatSoft,Tulsa, OK).

Results

Patients

The average age, gender, smoking and alcohol consumptionhabits of patients provide a good representation of the HungarianHNSCC population (Table 1). The mean age of the patients was53.7 years; the mean survival was 40.4 months. The majority of tu-mor samples were grade II, and the patients’ survival correlatedwith the grade (data not shown). Since the overwhelming majorityof patients were heavy smokers, we could not perform statisticalanalysis with regard to smoking.

EGFR gene copy number is frequently increased in HNSCC andcorrelates with poor survival

EGFR gene copy number was determined by FISH. EGFR geneamplification was found in eight of the 71 primary tumors, i.e. in11.2% of the tumors (Fig. 1A–B). Interestingly, it was more preva-lent in hypopharyngeal tumors (6 of 19) than in other locations.We found only one amplified sample in the supraglottic and onein the glottic tumors, and none in the base of tongue or tonsillarlocations. We detected polysomy in six cases (8.5%, the majorityof them displayed four centromeres) and trisomy in 25 patients(35.2%). Altogether 43.7% of the cases had increased EGFR genecopy number. Comparing the survival of patient groups with high-er EGFR gene copy number (amplification + polysomy) and normalcopy number, we found that the increased number of EGFR genewas significantly associated with poor survival (p = 0.041;Fig. 1C). Moreover, EGFR gene amplification and increased copynumber showed correlation with protein expression: in the tumorsamples with a higher number of copies the ratio of cells with 3+staining (using the antibody recognizing the intracellular domainof EGFR) was significantly higher (p = 0.02; Fig. 1D). However,using the antibodies against the extracellular domains of EGFRwe did not find any correlation between EGFR gene copy numberand protein expression (data not shown).

EGFR protein expression correlates with survival in HNSCC

EGFR protein expression and activation showed significant dif-ferences according to tumor sites (Table 2). Importantly, the vari-ous epitopes demonstrated distinct staining intensities (Table 2).Furthermore, high intratumoral heterogeneity was observed(Fig. 2A–D). In order to investigate the clinical significance of EGFRprotein expression we used Kaplan–Meier analysis to detect differ-ences in the survival of patients with increased EGFR proteinexpression in tumors compared with those of normal or lower le-vel. We found that increased EGFR protein expression (3+ staining,using the antibody against the intracellular domain) was associ-ated with a significantly (p = 0.042) poorer survival (Fig. 2E). Inthe case of the antibodies recognizing other domains of EGFR pro-tein no significant correlation was revealed. Interestingly, patientswhose tumors expressed increased activity of EGFR (detected byphospho-EGFR-specific antibody) showed a tendency for bettersurvival compared to patients with normal or decreased activationof EGFR, although the difference did not reach statistical signifi-cance (p = 0.14; Fig. 2F).

Frequent expression of vIII mutant form of EGFR in HNSCC

EGFRvIII mutation was detected with immunohistochemicalanalysis on TMA sections of paraffin-embedded HNSCC tumor tis-sue using L8A4 monoclonal antibody (Fig. 3A). Fifteen of 71 HNSCCsamples (21%) were positive for this alteration of EGFR. Most of

Figure 1 Detection of EGFR gene copy number and the correlation of protein expression with gene copy number. Normal (A) and increased copy number (B) of the EGFR gene(chromosome 7 green and EGFR red signal). (C) Kaplan–Meier survival analysis indicates decreased overall survival for patients with increased copy number of EGFR gene(amplification or polysomy) compared to patients with normal copy number. (D) High protein expression determined by immunohistochemical detection of the intracellulardomain of EGFR correlated with increased gene copy number. (For interpretation of the references to color in this figure legend, the reader is referred to the web version ofthis article.)

Table 2The extent of EGFR expression using anti-epitope antibodies in the different tumorsites.

Base oftongue

Tonsil Supraglottic Glottic Hypopharynx

n = 6 n = 10 n = 16 n = 20 n = 19

ECDLow 26.9 ± 23.2 31.8 ± 26.8 23.5 ± 31.9 38.7 ± 29.9 33 ± 39.3Normal 39.7 ± 19.7 59.2 ± 18.6 48.7 ± 25.6 51.5 ± 15.3 41.5 ± 22.6High 33.4 ± 33.7 9 ± 15.7 27.9 ± 29.7 9.8 ± 16.4 25.5 ± 31.2JMLow 17.7 ± 25.2 28 ± 50.5 41.7 ± 46.3 46.1 ± 37 30.5 ± 43.2Normal 59.7 ± 28.6 70.5 ± 33.5 44.3 ± 27.8 43.6 ± 18.4 50.5 ± 30High 22.6 ± 30.5 1.5 ± 4.6 14 ± 16.8 10.3 ± 10.4 19 ± 36.4ICLow 25.51 ± 38 37.5 ± 36.4 33.2 ± 19.3 46.5 ± 42.1 35.8 ± 38.2Normal 49.3 ± 27.2 52.1 ± 23.6 61.4 ± 18.7 45.1 ± 20.6 50 ± 23.2High 25.16 ± 24.9 10.4 ± 17.1 5.4 ± 9.4 8.35 ± 22.2 14.2 ± 17.9p-EGFRIncreased 3 4 8 4 6Low 3 6 8 16 13

Staining intensity defined as low (0 and 1+), normal (2+) and high (3+); data showpercentage of cells in the different intensity groups, mean ± SD. In the case of p-EGFR, data indicate the number of cases. ECD, extracellular domain; JM, juxtra-membrane domain; IC, intracellular domain; p-EGFR, phospho-EGFR. p-EGFRexpression was defined positive if staining was present in more than 25% of tumorcells.

B. Szabó et al. / Oral Oncology 47 (2011) 487–496 491

these mutations (8/15) affected the laryngeal region, but we foundthem in the tonsillar and hypopharyngeal tumors as well.Importantly, all vIII mutant cases displayed increased EGFR genecopy number in contrast to the non-mutant group where the

incidence of increased gene copy number is 30% (p < 0.0001).There was no difference in the level of EGFR activation or preva-lence of HPV infection between the mutant and wild-type group(Table 3).

No EGFR TK domain mutation was found in HNSCC

For the detection of mutations in exons 19 and 21 of the EGFRTK domain we used high resolution melting (HRM) analysis. The19del mutant HCC827 epithelial adenocarcinoma cell line andH1975 pulmonary adenocarcinoma line with a point mutation inexon 21 served as positive controls and H358 bronchoalveolarcarcinoma cell line expressing the wild-type EGFR TK as negativecontrol. In this cohort of 71 HNSCC cases we did not find anymutations in either exon 19 or exon 21 of EGFR (Fig. 3B).

Low incidence of KRAS codon 12 mutation in HNSCC

As KRAS mutation status is important for the treatment oftumors with anti-EGFR therapies, we determined KRAS codon 12mutations with RFLP analysis. PCR products from the genomicDNA were digested with BstnI restriction enzyme, which digeststhe wild-type sequence but not the mutant one, thus identifyingmutations in codon 12 (Fig. 4A). We quantified the ratio of themutant and wild-type products using Experion Automated Electro-phoresis System (Fig. 4B). Interestingly, one of the tumors carryingmutations had limited amount of wild-type allele present. Onlytwo samples of the 71 HNSCC cases contained the mutant allele(2.8%).

Figure 2 Immunohistochemical analysis of HNSCC samples using epitope-specific antibodies. Representative pictures show the different expression patterns of EGFRdomains and phosphorylation in tumor samples. Note the high intratumoral heterogeneity (A, C and D). (A) Note the nest of 3+ tumor cells in a 1+ homogenous tumor tissue.(B) The antibody specific to the juxtamembrane region demonstrated a gradually decreasing staining intensity in the normal skin (right side), the normal basal layer intensitywas designated as 2+. Tumor cells show a strong membrane-specific homogenous staining (left side). (C) The intracellular domain staining shows heterogenous signal in thiscase with larger tumor cells (1+) surrounded by smaller tumor cells displaying high staining intensity (3+). (D) Note the lack of EGFR activation in the connective tissue andthe varying level of activation in the tumor tissue. Scale bars = 50 lm. (EC: extracellular domain; JM: juxtamembrane domain; IC: intracellular domain) (E) Kaplan–Meiersurvival analysis shows significantly shorter survival for patients with 3+ tumors compared to those with lower expression level using antibody specific to the intracellulardomain. (F) High EGFR activation results in a somewhat longer survival, although the difference was not statistically significant.

492 B. Szabó et al. / Oral Oncology 47 (2011) 487–496

HNSCC patients with HPV-positive tumor did not demonstrate bettersurvival

We could demonstrate the presence of integrated HPV in a totalof 14 of the 71 cases (19.7%, Fig. 4C). Five tumors contained HPV16,six contained HPV18 and three tumors contained HPV33. Regard-ing tumor site, 33% of base of tongue, 30% of tonsillar, 19% of supra-glottic, 20% of glottic and 5% of hypopharyngeal tumors proved tobe HPV-positive, thus the infection incidence shows a decreasingtendency from the oral cavity towards the larynx. As opposed to

previous literature data, HPV-positive patients did not have bettersurvival rates in the studied population (Fig 4D). Interestingly,none of the positive samples showed EGFR gene amplification,while no statistically significant correlations were found betweenHPV infection, EGFR vIII mutation or EGFR activation (Table 3).

Prognostic factors in patients with HNSCC

Multivariate analysis of our data showed that in addition toclassical pathological parameters (such as vessel invasion and lym-

Figure 3 Detection of EGFR vIII and TK domain mutations in HNSCC samples. (A) Immunohistochemical analysis of EGFRvIII mutation using L8A4 antibody. Note the specificmembrane-localized staining. Scale bar = 25 lm. (B) Mutations of the TK domain in exons 19 and 21 were analyzed with high-resolution melting (HRM) technique. Positivecontrols were different cell lines with known mutations (grey lines). None of the HNSCC tumor samples (black lines) had mutations in the TK domain.

Table 3Comparative analysis of the various EGFR alterations.

Normal genecopy number

Increased genecopy number

p-value

vIII negative 38 16 0.0001vIII positive 0 15HPV negative 29 28 0.13HPV positive 9 3Vessel invasion negative 19 5 0.035Vessel invasion positive 16 15Surgical margin negative 32 9 0.049Surgical margins positive 15 12Inflammation No/mild 10 7 0.27InflammationModerate/strong 28 10

Low EGFRactivity

Increased EGFRactivity

vIII negative 35 21 0.77vIII positive 10 5HPV negative 37 20 0.71HPV positive 8 6Normal copy number 21 17 0.18Increased copy number 22 9Vessel invasion

negative6 17 0.035

Vessel invasionpositive

16 13

vIII neg. vIII pos.

HPV negative 45 12 0.1HPV positive 11 3

B. Szabó et al. / Oral Oncology 47 (2011) 487–496 493

phatic infiltration) the appearance of vIII-mutant form of EGFR pre-dicted outcome independent of other variables (p = 0.041, Table 4).

Interestingly, multivariate analysis also showed a similar ten-dency in the case of EGFR activity. However, this proved to be sta-tistically nonsignificant.

Lymph node metastases display altered EGFR expression andactivation

Currently one of the major questions regarding molecularly tar-geted therapies is whether primary tumors and their metastaseshave similar sensitivity to treatment. In order to investigate thisproblem we compared seven primary HNSSC tumors and theirlymph node metastasis (Table 5). Notably, altered gene copy num-ber was found in one pair and different vIII mutation status in an-other pair. Importantly, four metastases out of the seven pairsshowed decreased EGFR activity. We found no discrepancies inKRAS mutation status or in presence of HPV infection.

Discussion

The prognosis of HNSCC patients, especially in more advancedstages, is unfavorable. Therefore, new therapeutical approachesare needed for their treatment. Recently EGFR has become anattractive target in HNSCC, and anti-EGFR antibodies as well asTK inhibitors have been tested.24 In order to explore the expressionand genetic alterations of EGFR we processed samples from 71HNSCC patients, including five different tumor sites. The patientgroup is representative of the Hungarian HNSCC population: themajority of the patients are male, strong smokers and frequentalcohol consumers. Most of the tumors were stage II–III, and sur-vival was strongly related to tumor site and grade.

EGFR may have several alterations in tumors. In addition to pro-tein overexpression and gene amplification,25 TK domain muta-tion26 and deletion affecting the extracellular domain (the socalled vIII variant)27 were also described in tumors of various ori-gins. Nevertheless, KRAS, a key molecule in EGFR signaling andone of the most frequent oncogenes, can also be mutated28 andhas significant effects on the outcome of the anti-EGFR therapy.In our EGFR expression study on HNSCC we used four different epi-tope-specific antibodies covering the entire EGFR protein includingthe extracellular ligand-binding and membrane-proximal regionand the intracellular region. Furthermore, the activation of thereceptor was demonstrated by staining of the phosphorylatedreceptor in the same tumor samples. The staining intensity wasscored in comparison to the tumor-proximal normal epitheliumin each sample. Regardless of anatomical location, pronounced het-erogeneity was found in EGFR protein expression within the sametumor: a mixture of negative tumor cells and those that express arather high amount of the protein. It is noteworthy that EGFR over-expression (detected by the antibody recognizing the intracellulardomain of EGFR) was associated with significantly worse survivalrates in the patient cohort. This result confirms previous publica-tions indicating that EGFR protein overexpression is a negativeprognostic factor.7,29 However, using the antibodies against theextracellular region of EGFR did not recapitulate this finding. Thesedata suggest that the semiquantitative characterization of EGFRexpression by immunohistochemistry is rather difficult. This maypartly be due to the fact that the vIII mutant form of EGFR (com-plete deletion of the extracellular domain 2–7) does not react withthe extracellular domain antibodies. Importantly, a significantnumber of samples (21%) showed strong positive staining withthe antibody recognizing the vIII mutant form of EGFR.22 This re-sult is similar to earlier findings, where the rate of vIII-positivesamples was 42% in HNSCC patients.11 Nevertheless, a Japanesestudy could not detect any vIII mutant EGFR in HNSCC cases, whichmay reflect regional differences in the etiology of the disease.30 We

Figure 4 Detection of KRAS mutation and HPV infection in HNSCC samples. (A) Restriction fragments of the PCR products from HNSCC tumor samples. The uncut fragmentsderiving from the mutated gene migrate slower in the agarose gel (sample #15). (B) The ratio of wild-type and mutant alleles was determined by microcapillaryelectrophoresis of the fragments. Note the very low amount of the wild-type allele in this sample. (C) The presence of HPV genes was detected by PCR (lanes 2 and 3: HPV-positive samples, last lane: positive control). (D) Kaplan–Meier survival analysis demonstrated no difference between the patient groups with HPV-positive and HPV-negativetumors.

Table 4Multivariate analysis of the various EGFR alterations and pathological data asprognostic factors in patients with HNSCC.

Prognostic factor Relative risk (95% confidence interval) p

Surgical margin 1.036 (0.448�2.395) 0.935Vessel invasion 6.308 (2.171�18.327) 0.0007Lymphatic infiltration 3.561 (1.121�11.313) 0.031EGFR activity 0.518 (0.248�1.084) 0.081HPV 0.84 (0.352�2.007) 0.695vIII 2.88 (1.041�8.881) 0.041

Table 5Comparison of the primary tumors and their lymph node metastasis.

EGFRcopy

3+ ICdomain(%)

EGFRactivity

vIIImutation

KRASmutation

HPV

Hypopharynx Normal 35.60 Increased + wt �Meta Normal 0 Low � wt �Supraglottic Increased 0 Increased + wt �Meta Normal 0 Increased + wt �Glottic Normal 84.82 Increased � wt �Meta Normal 0 Low � wt �Tonsilla Normal 41.56 Low � wt �Meta Normal 0 Low � wt �Hypopharynx Normal 21.91 Increased + wt �Meta Normal 40.91 Low + wt �Hypopharynx Increased 29.54 Increased � wt �Meta Increased 0 Low � wt �Hypopharynx Increased 0 Low + wt �Meta Increased 0 Low + wt �

494 B. Szabó et al. / Oral Oncology 47 (2011) 487–496

found a strong correlation between EGFRvIII mutation andincreased gene copy number. However, we did not find any statis-tically significant correlation between the EGFR hyperactivation(detected by EGFR phosphorylation intensity) and gene amplifica-tion, HPV infection or the presence of vIII mutation. Interestingly,we found that the increased activity of EGFR showed a trend tocorrelate with longer survival. This suggests that in contrast toincreased gene copy number and protein level that have been asso-

ciated with poor prognosis, the increased activity of EGFR proteinseems to correlate with better prognosis. Similar correlations were

B. Szabó et al. / Oral Oncology 47 (2011) 487–496 495

described in lung cancer patients,31 but this is the first study dem-onstrating this association in HNSCC. Mutations of the TK domainof EGFR are frequent in some tumor types such as in non-small celllung cancer,26 but it is rare in HNSCC.10,15 A European study couldnot detect such a mutation, while a Japanese study found it in 7.3%of cases.14 Our results support the finding that EGFR-TK mutationmay be very infrequent in European HNSCC patients.

KRAS is the most frequently mutated oncogene, with mutatedalleles found in several tumor types in a large number of cases, of-ten up to 30–70%.32 However, these mutations are rather infre-quent in tumors of the head and neck region.8,17 In accordancewith previous findings we found only two patients with samplesexpressing the codon 12 mutant KRAS.

Recently the significance of HPV infection has been suggested forhead and neck tumors as well.20 The role of HPV infection has alsobeen considered as modulator of anti-EGFR therapies: HPV involve-ment has been found to be associated with better survival.19,33 Afrequent method to demonstrate the presence of high-risk HPV isthe PCR technique, but it is argued that in addition to the presenceof the gene, detection of p16INK4a protein expression is also a cri-terion of active HPV oncoprotein.34,35 Previous findings in the Hun-garian population showed 48.5% HPV positivity in cancers of theoral cavity and oropharyngeal regions and 35.7% in the case of lar-yngeal tumors.36 While all verrucous SCCs and most basaloid SCCswere HPV-positive, HPV infection could be detected only in 19.6% inthe typical SCCs which is in good correlation with our data showingthe presence of the most frequent high-risk HPVs (16, 18 and 33) ina total of 19.7% of our archived samples. However, we could notdemonstrate any statistically significant relationship with survival,EGFR protein expression, or other EGFR gene alterations. Neverthe-less, it was obvious that the incidence of HPV+ tumors shows adecreasing tendency as we proceed from the oral cavity towardsthe larynx, and there was no EGFR gene amplification in any ofthe HPV-positive samples. These results indicate that demonstrat-ing the evidence of HPV infection having any effect in HNSCC carci-nogenesis needs further investigation. However, it is worthmentioning that the majority (almost all) of the investigated pa-tients were heavy smokers and alcohol consumers so viral andchemical carcinogenesis could not be distinguished.

The different molecular characteristics of primary and meta-static lesions are critical factors in the outcome on molecularly tar-geted therapies. In line with this notion we studied seven pairs ofprimary tumors and their lymph node metastasis. We found evi-dence for different genetic alterations in two pairs suggesting theclonal origin of the metastatic tumors. Importantly the expressionand activation of EGFR was different in five cases suggesting thatthe primary and metastatic tumors may have different sensitivityto anti-EGFR therapies.

Our results have proven for a Hungarian HNSCC population thatof the potential alterations of EGFR in HNSCC only protein overex-pression and increased gene copy number may have prognosticsignificance. Since alterations described so far occur rather infre-quently, neither gene amplification, nor known mutations (EGFR-TK domain, KRAS) can play a significant role in tumor progressionin this localization. Based on this, target-specific anti-EGFR thera-pies will probably be insufficient in the case of HNSCC patientsand other potential targets ought to be identified. A recent studyhas suggested an important role for cMET (hepatocyte growth fac-tor receptor) in head and neck cancer development.37

The frequent incidence of deletion mutations (including vIIImutation) of the extracellular domain of EGFR in HNSCC will needto be taken into account for future planning of target therapies.Previous results11 and our findings here suggest that this mayundermine the outcome of currently used anti-EGFR antibodytherapies. It has been suggested that the extracellular deletionmay prevent antibody binding to the mutant EGFR protein, which

would lead to a decrease in therapeutic response.38 Further studiesare necessary to determine whether this mutant form interfereswith the efficacy of anti-EGFR antibody treatment.

Grant support

This work was supported by the following grants: National Sci-ence Foundation-OTKA-F68916; OTKA-NK73082 (B.D.); OTKA-K76293, OTKA-K-84173 (J.Tó.) and OTKA-NK-72595 (J.Ti.); EGT/Norwegian Financial Mechanism-HU0125 (B.D., J.Ti. and J.Tó.);ESMO Translational Research Fellowship (B.D.); RegIonCo-L00052Cross-border Co-operation Program Hungary, Austria 2007–2013(B.D., J.Tó.); NKFP_07_A2-NANODRUG OM-00080/2008 (I.P.); fur-ther support: J. Tóvári and B. Döme are recipients of the EötvösHungarian State Scholarship.

Conflict of interest statement

The authors declare that they have no competing financialinterests concerning this manuscript.

Abbreviations

EDTA

ethylenediaminetetraacetic acid EGF epidermal growth factor EGFR epidermal growth factor receptor FISH fluorescent in situ hybridization HNSCC head and neck squamous cell carcinoma HPV human papilloma virus PCR polymerase chain reaction PI propidium iodide SCC squamous cell carcinoma TK tyrosine kinase TKI tyrosine kinase inhibitor TMA tissue microarray

Acknowledgements

We kindly thank Katalin Derecskei, Violetta Piurkó, MagdolnaKanyik for their excellent technical assistance and Andrea Ladányiand Orsolya Szabo-Salfay for critically reviewing the manuscript.The authors express special thanks to Darell D. Bigner, Departmentof Neurology, Duke University, Durham, NC, for the anti-EGFRvIII(L8A4) antibody.

References

1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer JClin 2005;55(2):74–108.

2. Dobrossy L. Epidemiology of head and neck cancer: magnitude of the problem.Cancer Metastasis Rev 2005;24(1):9–17.

3. Goerner M, Seiwert TY, Sudhoff H. Molecular targeted therapies in head andneck cancer–an update of recent developments. Head Neck Oncol 2010;2:8.

4. Zimmermann M, Zouhair A, Azria D, Ozsahin M. The epidermal growth factorreceptor (EGFR) in head and neck cancer: its role and treatment implications.Radiat Oncol 2006;1:11.

5. Rocha-Lima CM, Soares HP, Raez LE, Singal R. EGFR targeting of solid tumors.Cancer Control 2007;14(3):295–304.

6. Herbst RS. Review of epidermal growth factor receptor biology. Int J RadiatOncol Biol Phys 2004;59(Suppl. 2):21–6.

7. Rogers SJ, Harrington KJ, Rhys-Evans P, OC P, Eccles SA. Biological significance ofc-erbB family oncogenes in head and neck cancer. Cancer Metastasis Rev2005;24(1):47–69.

8. Sheikh Ali MA, Gunduz M, Nagatsuka H, Gunduz E, Cengiz B, Fukushima K, et al.Expression and mutation analysis of epidermal growth factor receptor in headand neck squamous cell carcinoma. Cancer Sci 2008;99(8):1589–94.

9. Temam S, Kawaguchi H, El-Naggar AK, Jelinek J, Tang H, Liu DD, et al. Epidermalgrowth factor receptor copy number alterations correlate with poor clinicaloutcome in patients with head and neck squamous cancer. J Clin Oncol2007;25(16):2164–70.

496 B. Szabó et al. / Oral Oncology 47 (2011) 487–496

10. Chung CH, Ely K, McGavran L, Varella-Garcia M, Parker J, Parker N, et al.Increased epidermal growth factor receptor gene copy number is associatedwith poor prognosis in head and neck squamous cell carcinomas. J Clin Oncol2006;24(25):4170–6.

11. Sok JC, Coppelli FM, Thomas SM, Lango MN, Xi S, Hunt JL, et al. Mutantepidermal growth factor receptor (EGFRvIII) contributes to head and neckcancer growth and resistance to EGFR targeting. Clin Cancer Res2006;12(17):5064–73.

12. Yoshida T, Zhang G, Haura EB. Targeting epidermal growth factor receptor:central signaling kinase in lung cancer. Biochem Pharmacol 2010;80(5):613–23.

13. Lynch TJ, Bell DW, Sordella R, Gurubhagavatula S, Okimoto RA, Brannigan BW,et al. Activating mutations in the epidermal growth factor receptor underlyingresponsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med2004;350(21):2129–39.

14. Lee JW, Soung YH, Kim SY, Nam HK, Park WS, Nam SW, et al. Somatic mutationsof EGFR gene in squamous cell carcinoma of the head and neck. Clin Cancer Res2005;11(8):2879–82.

15. Lemos-Gonzalez Y, Paez de la Cadena M, Rodriguez-Berrocal FJ, Rodriguez-Pineiro AM, Pallas E, Valverde D. Absence of activating mutations in the EGFRkinase domain in Spanish head and neck cancer patients. Tumour Biol2007;28(5):273–9.

16. Carlson M, Wuertz B, Lin J, Taylor R, Ondrey F. Exons 19 and 21 of epidermalgrowth factor receptor are highly conserved in squamous cell cancer of thehead and neck. Int J Otolaryngol 2009;2009:649615.

17. Hoa M, Davis SL, Ames SJ, Spanjaard RA. Amplification of wild-type K-raspromotes growth of head and neck squamous cell carcinoma. Cancer Res2002;62(24):7154–6.

18. Perrone F, Suardi S, Pastore E, Casieri P, Orsenigo M, Caramuta S, et al.Molecular and cytogenetic subgroups of oropharyngeal squamous cellcarcinoma. Clin Cancer Res 2006;12(22):6643–51.

19. Goon PK, Stanley MA, Ebmeyer J, Steinstrasser L, Upile T, Jerjes W, et al. HeadNeck Oncol 2009;1:36.

20. Kreimer AR, Clifford GM, Boyle P, Franceschi S. Human papillomavirus types inhead and neck squamous cell carcinomas worldwide: a systematic review.Cancer Epidemiol Biomarkers Prev 2005;14(2):467–75.

21. Michl P, Pazdera J, Prochazka M, Pink R, Stosova T. Human papillomavirus inthe etiology of head and neck carcinomas. Biomed Pap Med Fac Univ PalackyOlomouc Czech Repub 2010;154(1):9–12.

22. Wikstrand CJ, Hale LP, Batra SK, Hill ML, Humphrey PA, Kurpad SN, et al.Monoclonal antibodies against EGFRvIII are tumor specific and react withbreast and lung carcinomas and malignant gliomas. Cancer Res1995;55(14):3140–8.

23. Fule T, Csapo Z, Mathe M, Tatrai P, Laszlo V, Papp Z, et al. Prognostic significanceof high-risk HPV status in advanced cervical cancers and pelvic lymph nodes.Gynecol Oncol 2006;100(3):570–8.

24. Bozec A, Peyrade F, Fischel JL, Milano G. Emerging molecular targeted therapiesin the treatment of head and neck cancer. Expert Opin Emerg Drugs2009;14(2):299–310.

25. Gupta R, Dastane AM, Forozan F, Riley-Portuguez A, Chung F, Lopategui J, et al.Evaluation of EGFR abnormalities in patients with pulmonary adenocarcinoma:the need to test neoplasms with more than one method. Mod Pathol2009;22(1):128–33.

26. Rosell R, Taron M, Reguart N, Isla D, Moran T. Epidermal growth factor receptoractivation: how exon 19 and 21 mutations changed our understanding of thepathway. Clin Cancer Res 2006;12(24):7222–31.

27. Mellinghoff IK, Wang MY, Vivanco I, Haas-Kogan DA, Zhu S, Dia EQ, et al.Molecular determinants of the response of glioblastomas to EGFR kinaseinhibitors. N Engl J Med 2005;353(19):2012–24.

28. Saridaki Z, Georgoulias V, Souglakos J. Mechanisms of resistance to anti-EGFRmonoclonal antibody treatment in metastatic colorectal cancer. World JGastroenterol 2010;16(10):1177–87.

29. Rubin Grandis J, Melhem MF, Gooding WE, Day R, Holst VA, Wagener MM, et al.Levels of TGF-alpha and EGFR protein in head and neck squamous cellcarcinoma and patient survival. J Natl Cancer Inst 1998;90(11):824–32.

30. Hama T, Yuza Y, Saito Y, Ou J, Kondo S, Okabe M, et al. Prognostic significance ofepidermal growth factor receptor phosphorylation and mutation in head andneck squamous cell carcinoma. Oncologist 2009;14(9):900–8.

31. Endoh H, Ishibashi Y, Yamaki E, Yoshida T, Yajima T, Kimura H, et al.Immunohistochemical analysis of phosphorylated epidermal growth factorreceptor might provide a surrogate marker of EGFR mutation. Lung Cancer2009;63(2):241–6.

32. Harris TJ, McCormick F. The molecular pathology of cancer. Nat Rev Clin Oncol2010;7(5):251–65.

33. Fakhry C, Westra WH, Li S, Cmelak A, Ridge JA, Pinto H, et al. Improved survivalof patients with human papillomavirus-positive head and neck squamous cellcarcinoma in a prospective clinical trial. J Natl Cancer Inst 2008;100(4):261–9.

34. Cuschieri K, Wentzensen N. Human papillomavirus mRNA and p16 detection asbiomarkers for the improved diagnosis of cervical neoplasia. Cancer EpidemiolBiomarkers Prev 2008;17(10):2536–45.

35. Deschoolmeester V, Van Marck V, Baay M, Weyn C, Vermeulen P, Van Marck E,et al. BMC Cancer 2010;10:117.

36. Szentirmay Z, Polus K, Tamas L, Szentkuti G, Kurcsics J, Csernak E, et al. Humanpapillomavirus in head and neck cancer: molecular biology andclinicopathological correlations. Cancer Metastasis Rev 2005;24(1):19–34.

37. Seiwert TY, Jagadeeswaran R, Faoro L, Janamanchi V, Nallasura V, El Dinali M,et al. The MET receptor tyrosine kinase is a potential novel therapeutic targetfor head and neck squamous cell carcinoma. Cancer Res 2009;69(7):3021–31.

38. Jutten B, Dubois L, Li Y, Aerts H, Wouters BG, Lambin P, et al. Binding ofcetuximab to the EGFRvIII deletion mutant and its biological consequences inmalignant glioma cells. Radiother Oncol 2009;92(3):393–8.